Machine Learning Approaches to Improve Serious Illness Communication among Patients with Advanced Cancer
The Serious Illness Conversation (SIC) Guide is a structured checklist for clinicians to elicit decision-making preferences, share patient-centric prognostic information, and explore trade-offs in serious illnesses. In clinical trials among patients with advanced cancer, eligibility criteria for SICs rely on physician-assessed one-year life expectancy, which is often inaccurate and under-identifies patients who may benefit. Recent advances in electronic health record (EHR) infrastructure, predictive analytics, and patient-reported outcome (PRO) collection may allow better identification of patients with high symptom burden or one-year mortality risk who would benefit most from an SIC. The long-term goal of my research program is to design and evaluate technology-focused interventions to improve palliative care delivery for patients with cancer. In this career development award, I propose to develop an algorithm-based intervention based on routinely collected EHR and PRO data to identify patients with cancer who may benefit from early SICs. The specific aims of this proposal are to (1) Explore desired characteristics of predictive algorithms to prompt SICs, using semi-structured interviews with oncology clinicians; (2) Compare characteristics of patients with high predicted mortality and/or patient-reported symptom burden to patients without these characteristics, using a prospective cohort study of 50 patients with advanced non-small cell lung (NSCLC) and gastrointestinal (GI) cancers; and (3) Test the impact of a clinician-centered intervention to prompt SICs using information regarding PROs and predicted one-year mortality, using a randomized controlled trial among 100 patients with advanced NSCLC and GI cancers. This proposal directly addresses a key NPCRC priority research area by exploring new methods to improve communication between patients with advanced cancer and their clinicians. This research will establish a generalizable approach to using advanced analytics and PROs to target supportive care interventions. My findings will provide preliminary data to support future efforts to develop technology-enabled interventions to improve palliative care delivery.
B. Parikh, MD, MPP, is an Instructor in Medical Ethics and Health Policy at the
University of Pennsylvania and a Staff Physician at the Corporal Michael J.
Crescenz VA Medical Center. Dr. Parikh is a practicing oncologist with
expertise in delivery system reform and informatics. His work has focused on
two core areas: (1) the use of predictive analytics to improve routine patient
care, particularly for those with advanced illnesses such as cancer, and (2)
quality of life and survivorship care among individuals with cancer. He
specializes in observational data analysis and pragmatic clinical trials.